Re: Motion P1788/M0050:EDP-Without-CA: YES
On 2013-09-16 14:32:35 +0300, Svetoslav Markov wrote:
> Here are my considerations:
>
> To the agument of Richard Fateman:
>
> I think that translation from end-point to mid-rad
> form and vice versa will benefit from the EDP. In particular,
> if one of the two forms is a number (degenerate interval),
> obtained as result of some computations, the other form
> may be necessarily nondegenerate interval. The latter will
> be sharper if an EDP is available.
There's no reason why EDP would improve inf-sup to mid-rad conversion.
Moreover an implementation is free to use whatever method it likes.
There is no need to specify additional operations for internals.
For instance, some form of multiple precision is useful (actually
needed) to implement correctly-rounded math functions (in IEEE 754,
and for tightest mode here in P1788). Does this mean that multiple
precision should be specified by the standard (IEEE 754 or P1788)?
The answer is no.
> To the agument of Vincent Lefevre:
>
> If EDP is available then the operations exact sum or an exact product
> of two numbers are readily available as special cases of EDP.
But they would still need to be specified. Just like though we
have the fma operation, we still want to specify addition and
multiplication, even though they are special cases of fma.
--
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)